From folder to gallery in one flow

In my last posts I shared how a small set of scripts grew into a working photo pipeline for renaming, resizing, watermarking, metadata and upload.
Since then the system crossed a line for me.
It now matches how I actually shoot and publish.
Today I am rolling out three changes that make the biggest difference in day-to-day work.

• Multi Set lets me queue several subjects at once. One click starts the whole batch.
• Review with a slider gives a quick way to align the quality rating when I approve images.
• Prefill subject caption and keywords uses Ollama with LLaVA to draft the text next to each image.

Multi Set saves the clicks

Sometimes a subject is a full series. Sometimes it is a single keeper.
Before this change I had to run the process each time.
Now I add sets into the queue in one window choose location and folder and press Start Batch.
The app picks up each set and moves through import extract scoring and staging for review without me babysitting it.
It sounds small but it removes the overhead between creative moments.
I can queue a bird, a skyline, and a single street shot in one go then let the system prepare everything for review.

Review with confidence

The Review Editor is where decisions are made.
I can zoom the image read the EXIF and with the QR slider give a quick verdict.
The models still run in the background to inform me but I do not need to think about numbers.
One slider sets a clear rating.

When a photo is approved, I can nudge the file name or keywords then press Publish.
The uploader handles the rest.
In the coots example six images went up in one batch with zero failures, captions and keywords included.

Prefill that saves typing

I added a Prefill step that uses Ollama with LLaVA to draft the subject caption and keywords.
It is local and fast.
Most of the time I keep the suggestion as it is.
When I want to be specific, I edit a word or two and move on.
The goal is simple.
Less manual text entry and more time shooting.

Live on the site with full context

Once published the image appears in the gallery with EXIF, description, and keywords.
Search works better and the archive stays consistent without extra effort.

Time saved and a clearer flow

• Fewer restarts thanks to Multi Set. Queue once and process many.
• Faster reviews with the slider. One gesture sets the quality rating.
• Better captions from Prefill. I edit only when needed.

On small sets the gain is a few minutes.
On travel days it is the difference between a late night of admin and a quick publish before bed.

What is next

I am working on two areas.
More small ML helpers that understand my style and sample testing to check a few frames from a fresh shoot before running the full batch.
The idea is to keep the system helpful without getting in the way.
Improving the image naming, caption, and keywords.

If you enjoy the build notes you can browse the code and the workflow here:

• Project on GitHub
• Workflow notes
• Previous posts: the first pipeline and the major update

Amir
Photographer, Builder, Dreamer
amir2000.nl